skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Install with Tessl CLI
npx tessl i github:boisenoise/skills-collections --skill skill-creator94
Does it follow best practices?
Validation for skill structure
Evaluation results
100%
86%Setting Up Evaluation Infrastructure for a New Skill
Eval workspace structure and metadata
Workspace naming
0%
100%
Iteration subdirectory
0%
100%
Descriptive eval names
58%
100%
eval_metadata.json presence
0%
100%
eval_metadata eval_name field
0%
100%
eval_metadata required fields
25%
100%
evals.json location and schema
40%
100%
evals.json entry fields
0%
100%
No-skill baseline structure
0%
100%
Simultaneous run launch plan
7%
100%
Without context: $0.5632 · 1m 55s · 30 turns · 34 in / 7,722 out tokens
With context: $0.9188 · 2m 43s · 32 turns · 5,421 in / 10,360 out tokens
88%
20%Create a Skill for Automated Git Commit Message Generation
Skill authoring style and structure
Pushy trigger language
0%
100%
Trigger contexts in description
0%
100%
SKILL.md line count
100%
100%
Imperative instructions
100%
100%
WHY reasoning present
100%
100%
No bare all-caps commands
100%
100%
Valid YAML frontmatter
100%
100%
evals.json skill_name and array
50%
100%
evals.json entry fields
50%
100%
No assertions in initial evals
100%
0%
Without context: $0.5128 · 2m 12s · 21 turns · 26 in / 7,108 out tokens
With context: $0.6567 · 2m · 24 turns · 5,450 in / 5,663 out tokens
94%
12%Improve Skill Triggering Accuracy with Eval Queries
Trigger eval query generation
Exactly 20 queries
0%
100%
Correct JSON structure
100%
100%
Sufficient should-trigger queries
100%
100%
Sufficient should-not-trigger queries
100%
100%
Specific should-trigger queries
100%
100%
Near-miss should-not-trigger queries
100%
100%
No obviously irrelevant queries
100%
100%
Varied should-trigger phrasings
100%
100%
Casual or informal queries
62%
100%
Query length variety
50%
40%
Without context: $0.2562 · 1m 35s · 11 turns · 15 in / 5,202 out tokens
With context: $0.6345 · 2m 37s · 17 turns · 5,410 in / 7,515 out tokens
- Evaluated
- Agent
- Claude Code
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.